JavaScript is disabled, so all seven pages are shown one after another on this single page. Enable JavaScript to use the navigation menu.

Talk to our team Email us
Recognition

ZNV, a leading AIoT solutions provider

Serving telecom operators and large enterprises.

400+

Invention patents

AAAAA

First batch of 5A-rated AI enterprises

CMMI-5

CMMI Level 5 certification

Standards Contributor

China's first AI enterprise standard

Top 100

2025 AIoT Top 100 enterprises

National AI Project

Sci-Tech Innovation 2030 AI programme

Customers

Trusted by 200+ customers

ZNV has delivered AIoT products and services to over 200 telecom operators
and large enterprises, including the names below.

  • China Mobile
  • China Unicom
  • China Telecom
  • China Tower
  • Tencent Cloud
  • Baidu AI Cloud
  • China Southern Power Grid
  • Beijing Energy International
  • JSTI Group
  • Sinopec
  • State Grid
  • TBEA
  • CNOOC
  • CNPC
  • Tianma Microelectronics
  • Digital China
  • Shanghai Stock Exchange
  • Am茅rica M贸vil
  • City of Windhoek
  • Ethio Telecom
  • Etisalat
  • Airtel
  • MTN
  • ZamGrid
Contact us

Tell us the scenario. We will tell you what is feasible.

Share the industry, the systems involved and the constraints you cannot break. Our team comes back with a reference architecture, comparable customer cases and a pilot plan.

Sentosa_DSML
Sentosa_DSML

AI FOR EVERY TEAM
ZERO CODE REQUIRED

Drag-and-drop modeling to deployment, all in one platform.

300+ OPERATORS 路 FULL LIFECYCLE 路 CLOUD, EDGE & ON-PREM

View product
Sentosa_AgentCube
Sentosa_AgentCube

ENTERPRISE AGENTS
BUILT FOR PRODUCTION

Govern every agent from prototype to live operations.

DUAL-DOMAIN ROLLOUT 路 3-TIER SANDBOX 路 FULL AUDIT TRAIL

View product
ZNV AC2000
ZNV AC2000

COOLING THAT ADAPTS
ENERGY THAT DROPS

On-site AI keeps every rack within safe temperature limits.

EDGE AI ENGINE 路 BUILT-IN FSU 路 UP TO 20% LESS ENERGY

View product
Contact us

Tell us the scenario. We will tell you what is feasible.

Share the industry, the systems involved and the constraints you cannot break. Our team comes back with a reference architecture, comparable customer cases and a pilot plan.

Sentosa_DSML

Zero code,
AI productivity unlocked

Turn the data your business already has into models that actually run in production
鈥?without writing a line of code.
Ingestion, modelling, evaluation and deployment, all in one platform.

Sentosa_DSML workspace: a drag-and-drop canvas of data, model and evaluation nodes
300+Built-in standard operators
~80%Shorter development cycle
100+Enterprise customers
70+Countries and regions served
Why

Why is enterprise AI modelling so hard?

Companies accumulate data for years and still cannot get a model into production.
These six obstacles come up in almost every project.

01

Business teams cannot code, so requests queue

Domain knowledge and modelling skill rarely sit in the same person. A single request can wait weeks before anyone picks it up.

02

Data preparation stretches the schedule

Data preparation and feature engineering consume most of the timeline, squeezing the time actually left for modelling.

03

Models rarely reach production

A trained model stays on a laptop, still a long way from the production system it was built for.

04

Results stay scattered and leave with people

Models and scripts sit on individual machines. When the author leaves, the work leaves with them and the same problem is solved twice.

05

Production data cannot go to a public cloud

Data security is a red line. Core production data cannot leave the intranet, which rules out cloud-only modelling platforms.

06

A platform is bought, then left idle

The real fear is that a new platform will not connect to existing systems 鈥?and will end up unused.

Capabilities

One platform, from data to deployment

Not a collection of point features 鈥?every step of modelling lives in one workflow.

Drag-and-drop modelling

Wire operators together on a canvas to build a model. Business engineers can start on day one.

Data access and governance

Files, relational databases, Hive, HBase, Elasticsearch and Kafka, with traceable data views.

~300 standard operators

Data processing, feature engineering, machine learning, deep learning and knowledge graphs.

Automated modelling and tuning

Automatic data preparation, algorithm selection and hyper-parameter search, with a recommended model.

Evaluation and comparison

Task-specific metrics and visualised training runs make overfitting easy to spot.

Model marketplace and versioning

Models and operator flows are versioned, reusable and managed as company assets.

One-click service publishing

Publish an operator flow as a RESTful API or a Docker image and put it into service.

Interactive notebooks

Algorithm teams keep working in code, then register the result as an operator for everyone else.

Difference

How is it different?

What is scarce is a tool business engineers can use, that reaches production, and that is proven.

Drag-and-drop that genuinely lowers the bar

Not a half-finished tool for developers 鈥?a complete platform process, quality and equipment engineers can use directly.

~300 operators covering everyday enterprise tasks

Classification, regression, time series, anomaly detection, image and text; structured and unstructured data alike.

Training, evaluation, tuning and deployment in one loop

Data to model and model to service happen in a single platform instead of a chain of stitched-together tools.

Private, domestic-stack and edge deployment

Adapted to domestic CPUs and operating systems; models can be exported as images and pushed to machine rooms, sites and production lines.

Proven on large enterprise projects

Live projects in petrochemicals, energy and telecom, so buyers can look at how their own industry uses it.

Scenarios

Can DSML solve your problem?

Look at the scenarios already running, not the industries a platform claims to support.

Chemicals 路 process parameter optimisation

Predict yield and product properties, recommend the best parameter combination and cut hydrogen and energy consumption.

Energy 路 equipment fault diagnosis

Judge equipment anomalies and degradation trends from condition data 鈥?from experience-based inspection to early warning.

Telecom 路 machine room energy saving

Forecast consumption and optimise air-conditioning strategy; pilot sites improved overall electricity saving by 21%.

Manufacturing 路 root cause and quality prediction

Identify abnormal conditions and locate the root cause, then predict quality from process parameters and intervene early.

General 路 time-series forecasting and anomaly detection

Forecast output, energy, load and temperature, and surface anomalies and abnormal patterns in the data.

Universities 路 teaching and research modelling

Lab work for teaching and research, and joint laboratories between universities and industry.

Cases

Proven in production

Results from projects already running in customer environments, not from a slide deck.

Sinopec Chemicals · quality analysis

Sinopec Chuanwei Chemical — root-cause analysis of production quality swings

An AI-driven analysis platform built on Sentosa_DSML. Hundreds of process parameters from across the production line are linked, ranked by their influence on product quality and traced to cause, so the hidden factors that manual experience had missed become visible.

  • Multi-dimensional parameter correlation and causal inference
  • Automated discovery of hidden influencing factors
  • End-to-end diagnostic model: process parameters in, ranked causes out
4 months → 1 hourquality issue investigation
5hidden influencing factors identified
CNOOC Refining · no-code modelling

CNOOC Huizhou Petrochemical — no-code modelling for refining processes

A refining big-data AI modelling platform built on the enterprise industrial PaaS, on a CPU–GPU hybrid compute architecture. Data scientists and process engineers develop, train and deploy models on one drag-and-drop canvas instead of writing code.

  • One-stop development to deployment
  • 28 refining production scenarios modelled
  • Process knowledge held as reusable data and model assets
100+AI models built
~1 week → ~2 hoursper modelling cycle
Security

Enterprise-grade, data never leaves your site

Deployment options and governance are built for long-term operation inside an enterprise intranet.

Deployment options

  • Standalone or clustered distributed deployment, delivered as containers
  • Private cloud, on-premises machine room or offline intranet
  • Models exported as files or Docker images for edge deployment

Permissions and audit

  • Department, user and role permission model
  • Publishing goes through an approval centre with a full trail
  • Encrypted password storage, verified sign-in, periodic rotation

Domestic technology stack

  • Intel, Kunpeng and Phytium CPUs
  • Windows, Linux and Kylin operating systems
  • Ascend and other domestic AI accelerator cards
FAQ

Questions buyers ask before a pilot

What is Sentosa_DSML?

Sentosa_DSML is ZNV's own one-stop AI development and deployment platform. It supports both drag-and-drop, zero-code modelling and interactive notebooks, and takes enterprises, cities, universities and research institutes through ingestion, processing, feature engineering, modelling, evaluation, tuning and deployment.

Can someone who cannot code build a model?

Yes. Operator flows are the primary way to work: drag the data, cleaning, feature-engineering, modelling and evaluation operators onto the canvas and connect them. Domain knowledge is still required 鈥?you need to know what each field means and what you are predicting.

Which task types are supported?

Classification, regression, clustering, anomaly detection and time-series forecasting are standard. Image classification, object detection, text classification and information extraction are handled by the deep-learning capability, which supports custom convolutional networks and imported models trained in TensorFlow, PyTorch or PaddlePaddle.

Does it support automatic hyper-parameter tuning?

Yes. Automated modelling covers automatic data preparation, algorithm selection and hyper-parameter search, comparing candidates against the given data and objective, and recommending the best model.

Can it connect to our own data?

Yes 鈥?CSV and Excel files, NFS, JDBC relational databases, Hive, HBase, Elasticsearch and Kafka, plus object storage and APIs. Data can be exposed as governed data views with quality monitoring and full lineage, so nothing has to leave the site.

Does it support private deployment?

Yes. Standalone and clustered distributed deployment, delivered as containers (Docker and Kubernetes), running in the customer's private cloud, machine room or offline intranet.

Does it work in an offline intranet?

Yes. The private deployment runs on customer intranet servers or the industrial-internet platform, with no production data leaving the network. Dependency packages and images are supplied as offline installation bundles, and models and images can move between the intranet and edge environments.

How is a trained model deployed as a service?

Operator flows are packaged as services in the service marketplace and published as RESTful APIs. Models can also be exported as files, or packaged as Docker images and deployed to workstations, servers or edge devices.

Does it support edge deployment?

Yes. A published model service can be exported as an image with one click and deployed outside the platform, in Docker on a tower workstation or server. Typical scenarios include on-site inference in machine rooms and base stations, health analysis for photovoltaic sites, and quality and anomaly detection on industrial lines.

How is this different from Python or Jupyter?

Python and Jupyter require you to write and maintain the code, and the deliverable is code. DSML delivers managed model assets inside a governed platform, with versions, approvals, publishing and permissions, and can generate an API or an image in one click. The two are complementary: DSML also supports notebooks, so algorithm teams can keep writing code and register the result as an operator.

How is this different from AutoML?

AutoML solves automated algorithm selection and tuning. DSML covers a much wider span: data access and governance, feature engineering, machine learning and deep learning modelling, knowledge graphs, model management, service publishing and workflow scheduling. Automated modelling is one module inside it.

How do we arrange a demo or a pilot?

Contact us with your industry and the business problem you want to solve. A formal pilot typically runs two to four weeks, with modelling support and platform training provided by ZNV.

Contact us

Tell us the scenario. We will tell you what is feasible.

Share the industry, the systems involved and the constraints you cannot break. Our team comes back with a reference architecture, comparable customer cases and a pilot plan.

Sentosa_AgentCube

Enterprise agents,
engineered for production

An engineering foundation for enterprise AI agents 鈥?not for a chat demo.

Sentosa_AgentCube workspace: an agent workflow with input, LLM and output nodes
95%Recall on 10-million-document libraries
15sRetrieval across the same corpus
3-tierCommand allowlist, hook interception, MicroVM
40%+Lower operating cost
Why agents stall

Three problems stop agents at the pilot stage

The same three obstacles recur across nuclear power, grid, energy and emergency response.

01

Black boxes cannot be trusted with critical work

The reasoning is invisible: users see the output, not why it was produced, what tools were called or which knowledge was used. Without explainability there is no audit, no traceability and no trust.

02

Incomplete recall makes hard questions unanswerable

Enterprise knowledge bases run to millions of documents. Typical retrieval pipelines reach roughly 60% recall at that scale, so nearly four in ten relevant documents are never found.

03

Uncontrolled execution breaks compliance

Agents call tools, read sensitive data and run commands. Without execution boundaries and isolation they can act dangerously, cross permission lines or produce side effects that cannot be rolled back.

What critical industries ask for

  • ExplainableThe reasoning process is white-box: every step is visible and auditable.
  • Complete recallOver 95% recall on libraries of ten million documents.
  • ControllableExecution has boundaries, an audit trail and rollback.
  • EvolvableAgents improve independently, without retraining the model.
  • DeployablePrivate deployment; data never leaves the domain.
Design

Four principles behind the platform

Engineered

Development and deployment live in two isolated domains. A validated agent becomes a production workflow with one click 鈥?the rigour of traditional software, applied to agents.

White-box

A black-box agent's reasoning is converted into a visible, editable, auditable workflow. Every decision, tool call and knowledge reference is recorded and shown.

Controlled

Command allowlists, hook interception and MicroVM isolation bound every execution. Sensitive operations are intercepted and logged; the runtime is fully separated from the host.

Evolvable

Agents accumulate experience and refine strategy without retraining. Degradation detection and automatic rollback restore a stable version when quality drops.

Capabilities

Eight modules across the agent lifecycle

Together they take an agent from “usable” to “usable in a critical business process”.

Layered memory

Short-term session memory, mid-term task state and long-term experience, each with configurable capacity, retention and access rights.

Knowledge engineering

RAG, GraphRAG, a semantic layer and tagging, with a virtual directory tree 鈥?15-second retrieval across ten million documents.

Development-to-deployment conversion

A validated agent becomes a production workflow automatically, with environment isolation, versioning and canary release.

Agent white-boxing

The decision logic of a black-box agent is converted into a visible, traceable, editable workflow.

Three-tier sandbox

Command allowlist, hook interception and MicroVM isolation, with fine-grained permission control and a full audit trail.

Tool integration

MCP protocol tools, CLI, RESTful APIs and plug-ins, with registration, versioning, permissions and usage monitoring.

Audit and evaluation

Evaluation sets and automated testing during development; full-chain auditing and A/B comparison in operation.

Model and token governance

Task-aware model routing, token consumption monitoring and quotas by department, project or agent.

Knowledge engineering

Recall that holds up at ten-million-document scale

Conventional retrieval-augmented generation truncates documents; PageIndex keeps them whole.

Knowledge engineering compared
DimensionSentosa_AgentCubeTypical approach
Retrieval time, 10m documentsWithin 15 secondsMinutes
Recall, 10m documents95%~60%
Document handlingComplete, no truncationChunked and truncated
Knowledge organisationRAG, GraphRAG, semantic layer, tagging, virtual directory treeSingle RAG pipeline

Virtual directory tree

Documents are organised logically by plant, equipment, discipline, document type and version, narrowing every query to the relevant range.

Hybrid retrieval

Keyword search for exact matches, vector search for intent, graph pattern matching for relationships 鈥?combined with semantic de-duplication and dynamic ranking.

Evidence traceability

Every conclusion carries its source passage, document version, data link, similarity score and retrieval scope, so it can be verified.

Cases

Proven in critical infrastructure

Pumped-storage operators run these agents in production, inside their own networks.

China Southern Power Grid Pumped storage · O&M

CSG Pumped Storage — fault diagnosis and condition-based maintenance

An agent for pumped-storage stations, built on Sentosa_AgentCube over the customer's existing big-data platform, MIS and maintenance standards. A knowledge graph links every device to its fault knowledge, and a fine-tuned power-domain model supports FMEA digitalisation, fault diagnosis, condition-based maintenance and inspection planning.

  • Knowledge graph tying devices, faults and work history together
  • Fault patterns matched against historical cases and disposal records
  • Inspection plans and fault reports generated on demand
≥95%accuracy, knowledge Q&A
≥90%fault diagnosis accuracy
≥90%maintenance plan accuracy
Value

What changes for the business

Efficiency

Retrieval drops from minutes to 15 seconds and recall rises from around 60% to 95%. Fault diagnosis moves from hours to minutes, and the gap between building an agent and running it in production shrinks sharply.

Quality

White-box workflows and full-chain auditing turn agent quality from a subjective impression into measurable indicators, with complete evidence for regulators and customers.

Cost

Model routing and token governance keep spend under control, while asset reuse and autonomous evolution reduce redevelopment and retraining. Operating cost falls by more than 40%.

Compounding assets

Agents, workflows, skills and knowledge bases are managed as versioned assets. They stay with the company when people move, and gain value as they are used.

Contact us

Tell us the scenario. We will tell you what is feasible.

Share the industry, the systems involved and the constraints you cannot break. Our team comes back with a reference architecture, comparable customer cases and a pilot plan.

Edge Hardware Series

Edge hardware that keeps
field data working

Gateways, AI boxes and devices that collect, analyse and act on data
where it is produced — inside the machine room, the site or the cabinet.

ZNV AC2000

ZNV AC2000

AI group-control gateway for air conditioners

Combines IoT access with an AI-driven group-control energy-saving unit. One gateway covers power and environment monitoring and adaptive air-conditioning control.

  • IoT device access and control
  • Data storage and analytics
  • Online inference and model training
  • Visual energy monitoring

Typical deployment Data centres · Government and public buildings · Enterprise comms rooms

ZNV AX2000-S320

ZNV AX2000-S320

High-density video analysis AI box

32 TOPS of INT8 AI computing power for real-time structured analysis of live video at the sites that carry the most camera traffic.

  • 32 TOPS of INT8 AI computing power
  • 20 real-time algorithms or 200 non-real-time algorithms
  • Real-time identification of people, vehicles, behaviour and objects
  • Integrates with VMS platforms, cameras and video walls
ZNV AX2000-J060

ZNV AX2000-J060

Compact video analysis AI box

The same structured video analysis in a smaller enclosure, sized for branch sites, substations and lighter camera loads.

  • 6 TOPS of INT8 AI computing power
  • 5 real-time algorithms or 20 non-real-time algorithms
  • Works with VMS, cameras and video walls
ZNV IG2200

ZNV IG2200

Integrated gateway

The field-side integration point of the ZNV AIoT hardware portfolio: IP and serial access, digital and analogue I/O, and local processing in a 1U chassis.

  • LAN, COM, USB, DI/DO and 12V/6V I/O in one unit
  • Runs edge applications locally, no cloud round trip
  • Rack-mount 1U form factor for machine rooms
ZNV Smart Lock and Key

ZNV Smart Lock & Key

One key, total control, maximum security

Digital-encryption locks and keys for cabinets, racks and access points. No power supply, no maintenance, and a full event record.

  • One key opens every authorised lock
  • Digital encryption — the key cannot be copied
  • Bluetooth app authorisation
  • Full audit trail of every event
Which unit

Five units, five jobs

Each unit above is built for one job.
If your site sits between two rows, tell us about it and we will pick one.

Unit The job it does Where it goes
ZNV AC2000 Air-conditioning group control and power and environment monitoring in one gateway Data centres, government and public buildings, enterprise comms rooms
ZNV AX2000-S320 32 TOPS of INT8 computing for real-time structured analysis of heavy video traffic Sites carrying the most cameras
ZNV AX2000-J060 The same structured video analysis in a compact enclosure Branch sites, substations and lighter camera loads
ZNV IG2200 Field-side integration: IP and serial access, digital and analogue I/O, local processing in 1U Machine rooms, rack-mounted
ZNV Smart Lock & Key Digitally encrypted locks and keys — no power supply, no maintenance, full event record Cabinets, racks and access points
Contact us

Tell us the scenario. We will tell you what is feasible.

Share the industry, the systems involved and the constraints you cannot break. Our team comes back with a reference architecture, comparable customer cases and a pilot plan.

AI Engineering Services

From data foundations to deployed agents

Three services that take an enterprise from scattered data to working AI:
govern the data, engineer the knowledge, and build the agents that run the business.

01

Data governance

The prerequisite for data intelligence. Clean, well-described data is what makes every model and every agent downstream usable.

Value-chain data governance

Governance of the data that runs through the enterprise value chain 鈥?sources, definitions, quality and lineage.

Corpus processing for large models

Preparation and processing of the corpora that large models are trained and grounded on.

City-scale data governance

Full-domain, end-to-end data governance for city-level programmes.

  • DCMM-3 data management maturity
  • Shenzhen Data Exchange registered data provider
02

Knowledge engineering & model customisation

Built on our own AI engineering toolchain and general-purpose model platforms,
this service delivers data and knowledge engineering across industries and domains:
how knowledge is acquired, represented, stored and retrieved,
and how models are customised, evaluated, deployed and applied.

Knowledge processing

Document processing and knowledge layering

  • Batch ingestion and tagging of PDF, Word and Excel material
  • A three-level hierarchy 鈥?domain, sub-domain, knowledge point 鈥?matched to the business scenario
Knowledge augmentation

Extraction, Q&A construction and structuring

  • Entity recognition and relationship triple extraction
  • Context-based question鈥揳nswer pair generation for QA augmentation
  • Structured augmentation: cleaning, triple extraction and feedback-driven refinement
Model customisation

Custom models and retrieval optimisation

  • Custom machine-learning and deep-learning model development, plus industry model delivery
  • Embedding model fine-tuning with retrieval and re-ranking enhancement
Semantic layer

Data dictionary and semantic modelling

  • Standardised data dictionary governance and metadata management
  • Semantic models covering entities, relationships, metrics, dimensions, joins and caching
  • Tencent Cloud algorithm model development partner
  • Baidu AI Cloud algorithm model development partner
  • Tencent Cloud authorised service partner (ASP)
  • Works with DeepSeek, ERNIE and Tencent Hunyuan
03

Agent development

Backed by multi-industry delivery experience, ecosystem partnerships
with general-purpose model vendors, and twenty years of R&D management discipline,
we build high-quality agent applications that fit your own business scenarios 鈥?br>and help you move to AI-enabled operations fast.

Development modes

Multi-mode agent development

  • Standard mode 鈥?retrieval-augmented Q&A agents
  • Workflow mode 鈥?highly customised SOP design and execution
  • Multi-agent mode 鈥?hybrid orchestration of autonomous planning agents
Data

High-quality data construction

  • Data augmentation: datasets plus business knowledge and Q&A pairs to improve model comprehension
  • Data generation: scenario-ready SFT and DPO training sets for fine-tuning
Integration

MCP plugin integration

  • Built-in MCP plugins give models access to external data sources and tools
  • Custom plugins configured over MCP to extend what an agent can reach

End-to-end delivery

Requirements, design, build, testing, deployment, operations and training.

Custom development

Integrated with existing systems such as WeCom and DingTalk; bespoke modules on top of agent primitives.

Engineering depth

CMMI Level 5 certified, IPD-based development process, 20+ years of R&D management.

Proven practice

Government, industry, energy and cultural-sector projects, with the risks already mapped.

  • Baidu AI Cloud large-model service innovation partner
  • Tencent Cloud certified service partner (CSP)
  • Delivered on Tencent Cloud, Alibaba Cloud Bailian, Coze and Baidu Qianfan AppBuilder
Contact us

Tell us the scenario. We will tell you what is feasible.

Share the industry, the systems involved and the constraints you cannot break. Our team comes back with a reference architecture, comparable customer cases and a pilot plan.

About ZNV

Shenzhen ZNV Technology Co., Ltd.

A leading provider of AIoT products and services, serving telecom operators and large enterprises.

ZNV employees photographed from above at a company anniversary event
Who we are

AIoT for the systems that cannot stop

For telecom operators, ZNV provides intelligent edge hardware, AIoT systems for power and environmental monitoring, intelligent operations and maintenance (O&M), energy efficiency, and computing–energy coordination. For large enterprises, ZNV offers enterprise-grade AI platforms, AIoT systems for complex business scenarios, and AI engineering services.

ZNV is among China's first 5A-rated AI enterprises, has undertaken a major project under the national "Science and Technology Innovation 2030 – New Generation Artificial Intelligence" initiative, and was named to the "2025 AIoT Top 100 Enterprises" list.

To date, ZNV has delivered AIoT products and services to over 200 telecom operators and large enterprises, including China Mobile, China Unicom, China Telecom, China Tower, Tencent Cloud, Baidu AI Cloud, China Southern Power Grid, Beijing Energy International, CGN, JSTI Group and Sinopec.

We build for telecom operators, energy groups and large enterprises — the organisations whose operations run around the clock and whose data cannot leave the site. That shapes everything we ship: edge hardware that works offline, platforms that run on private and domestic infrastructure, and agents that can be audited.

Recognition

ZNV, a leading AIoT solutions provider

Serving telecom operators and large enterprises.

400+

Invention patents

AAAAA

First batch of 5A-rated AI enterprises

CMMI-5

CMMI Level 5 certification

Standards Contributor

China's first AI enterprise standard

Top 100

2025 AIoT Top 100 enterprises

National AI Project

Sci-Tech Innovation 2030 AI programme

Customers

Trusted by 200+ customers

ZNV has delivered AIoT products and services to over 200 telecom operators
and large enterprises, including the names below.

  • China Mobile
  • China Unicom
  • China Telecom
  • China Tower
  • Tencent Cloud
  • Baidu AI Cloud
  • China Southern Power Grid
  • Beijing Energy International
  • JSTI Group
  • Sinopec
  • State Grid
  • TBEA
  • CNOOC
  • CNPC
  • Tianma Microelectronics
  • Digital China
  • Shanghai Stock Exchange
  • Am茅rica M贸vil
  • City of Windhoek
  • Ethio Telecom
  • Etisalat
  • Airtel
  • MTN
  • ZamGrid
Contact us

Tell us the scenario. We will tell you what is feasible.

Share the industry, the systems involved and the constraints you cannot break. Our team comes back with a reference architecture, comparable customer cases and a pilot plan.